Current limitations | When conducting a new study |
Heterogeneity in variable selection, definition and measurement, study design and statistical analyses | Consider previous similar studies when selecting determinants and behaviours |
Clarify variable definitions in relation to previous studies | |
Consider using established measures of adherence behaviours and determinants if available | |
Consider using established study designs and data analysis methods if appropriate | |
Limited theoretical basis for variable selection and lack of an integrated theoretical approach | Use existing behavioural theory to select variables Focus on testing multi-determinant models instead of a few preferred determinants If testing new models, clarify the choice and relationships with existing theories |
Lack of robust study designs for causal inferences in most studies | Prioritise the use of repeated measure longitudinal designs Assess adherence determinants prior to behaviours Choose time lags in which causal influence is likely Control for other possible causal influences |
Low or medium quality participant selection in some studies | Use prior literature to decide on clear inclusion criteria that allow comparisons with other studies |
Employ systematic procedures for participant selection | |
Report participant selection procedures clearly and completely | |
Insufficient description of variable definitions and measurement | Provide a clear rationale and description for included variables Provide comprehensive descriptions of measurement tools or methods in the manuscript or supplementary materials |
Low quality of measurement | Select or develop psychometrically sound measures |
Examine psychometrics as preliminary analyses | |
Report results of psychometric evaluation | |
Sources of bias rarely addressed | Reflect on possible sources of bias (e.g. response, recall, surveillance bias) and take steps to minimise their effect |
Study size rarely addressed | Consider the probability of type I and type II errors given the research question, population and resources available |
Low or medium quality data analysis procedures in most studies | Consult methodological literature relevant for the intended analyses Perform and report on preparatory analyses (e.g. missing data) Do not group continuous data unless solid justification exists and analyses are performed with both continuous and grouped data Control for possible confounders and justify their selection Adjust for sampling strategy and hierarchical data structures |